Opinion: Artificial 'Intelligence': Unemployment system denied legitimate COVID-19 claims

Jennifer Lord

The flood of coronavirus-related claims for unemployment benefits caused some 50% of the country’s state unemployment agency websites to crash. The millions of legitimate and fraudulent claims overwhelmed the computer systems used by most states to administrate unemployment benefits.

Michigan was no exception, as the state’s Unemployment Insurance Agency (UIA) reported that high volumes of claims filed temporarily crashed its website and resulted in jammed phone lines preventing some claimants from getting through for weeks.

Michigan’s seasonally adjusted unemployment rate advanced by
eight-tenths of a percentage point to 6.9% in November

The UIA’s controversial MiDAS (Michigan Integrated Data Automated System) computer system flagged over 540,000 (of 1.7 million) claims as being possibly fraudulent, and took the step of freezing the accounts and suspending payment of benefits to ferret out false and fraudulent claims. Unfortunately, the hold-up in benefits created hardships for tens of thousands of legitimate claimants.

Adding insult to injury, thousands of state residents whose accounts had been inaccurately flagged as fraudulent found it virtually impossible to successfully upload identity verification documents to prove they were who they said they were and re-establish their legitimate claims. 

The stories being told by some of the thousands of people whose benefits were cut off and falsely accused of filing fraudulent claims sounded familiar to me because I had heard them before.

Since 2015, our law firm has been representing some 40,000 state residents in a class action lawsuit alleging that UIA’s MiDAS system had similarly and erroneously accused them of fraud and billed them for repayment of benefits plus huge fines, penalties and interest.

Many of those we are representing were also victimized by aggressive collection practices, including wage garnishment, bank levies and seizure of income tax refunds. In addition, a significant percentage of them have experienced a range of long-term problems, such as personal bankruptcy, credit rating problems, eviction and divorce.

The State of Michigan has fought this class action for over five years. The case is currently being considered by the Michigan Supreme Court.

Many of the problems which paralyzed the UIA during the first months of the Pandemic Unemployment Assistance program were due in part to pre-existing problems with its artificial intelligence-based algorithm, which too often rendered erroneous determinations.

Despite ample evidence that the MiDAS system continues to underperform and bears partial blame for customer service meltdowns at UIA, the state appears fully committed to continuing its contract with FAST Enterprises, the vendor which provides MiDAS. In fact, the State of Michigan has recently extended the agreement with FAST Enterprises through the end of the year.

Increased Use of AI by Governments Inevitable

In defense of Michigan and other states, multiple factors — including budget pressure, reduced head counts and continuing emphasis on more efficiency in delivering services and benefits — will continue to drive the need for AI-based technologies which can do the work of humans for a fraction of the cost.

A long list of state CIOs and IT officials are convinced that AI is the transformational force required to streamline the functions of 21st century state governments. 

There is little doubt that AI has the potential to be a game-changer in delivering faster and more efficient administration of massive bureaucracies, especially as it can be utilized to deliver critical monetary, health care and related social benefits for millions of low-income, underprivileged residents. Government agencies on all levels have begun to utilize it for this purpose.

Many states, including Michigan, see solutions to customer service problems in AI-based chatbots and digital assistants that can deliver the support needed for over-extended help desks.

The theory — and the hope — is that digital technologies will lighten the load of responding to routine customer inquiries, thus freeing-up human employees to devote more time and resources to other, more important functions.

Vetting, Transparency Critical

Numerous state officials from throughout the U.S. see AI making the biggest impact in cybersecurity; fraud, waste and abuse detection and management; and improving citizen-facing digital services.

President Donald Trump and others have suggested that there is rampant fraud in government welfare programs. Pledges to crack down on the fraud and waste alleged in America’s social safety net has come from both conservative and liberal office-seekers.

There is scant evidence to back up this claim of rampant benefits fraud and cheating. For example, the country’s food stamp program, the Supplemental Nutrition Assistance Program (SNAP), serves about 40 million people a month at an annual cost of $68 billion as of 2018. Yet, the federal government itself says less than 1% of benefits go to ineligible households. Nationally, estimated intentional fraud rate in unemployment programs is equally negligible.

Unfortunately, dozens of states are taking these mostly unfounded accusations seriously and turning to artificial intelligence and other automated systems to determine benefits eligibility and ferret out fraud in a range of aid programs, from food stamps and Medicaid to unemployment insurance.

Prior experience with AI-based system has proven that many algorithms exhibit racial and socioeconomic bias, much like humans. For those of us who’ve seen the damage that a biased algorithm can do to a vulnerable individual or family, it’s clear that AI-based decision-making systems must be carefully vetted to ensure that they are not violating the legal and human rights of those applying for benefits.

The State of Michigan should follow the lead of Alabama, California, Vermont, New York City and the European Union and establish an AI task force to ensure that its automated decision-making systems are transparent, unbiased and ethical.

Jennifer Lord is a partner at Pitt, McGehee, Palmer, Bonnani & Rivers. She practices civil rights and employment law. She is part of the legal team in Bauserman vs. Unemployment Insurance Agency, a class action seeking compensation for 40,000 Michiganians falsely accused of fraud by the Unemployment Insurance Agency. That case is pending in the Michigan Supreme Court.